We will use Grok 3.5 (maybe we should call it 4), which has advanced reasoning, to rewrite the entire corpus of human knowledge, adding missing information and deleting errors.
Then retrain on that.
Far too much garbage in any foundation model trained on uncorrected data.
An interesting thought experiment: I think he’s full of shit, you think he’s full of himself. Maybe there’s a “theory of everything” here somewhere. E = shit squared?
He is a little shit, he’s full of shit, ergo he’s full of himself